Applied Intuition Details Sensor Stack for Autonomous Haul Trucks
Autonomous haul trucks use lidar, camera, and radar together to navigate unstructured mine and construction sites without lane lines or signage.
- 99%
- 1%
What Happened
Autonomous haul trucks operate in extremely unstructured environments like mines and construction sites, where there are no lane lines, curbs, or directional signs. Safe navigation relies on perception, and because no two sites are alike, a universal sensor solution doesn't exist. Applied Intuition's sensor stack uses lidar, camera, and radar in parallel to fill each other's gaps.
Lidar provides high-resolution 360-degree coverage but can fail on certain materials. Cameras enrich the model's understanding beyond lidar intensity alone, though they struggle in low light. Radar detects dynamic obstacles via the Doppler effect, providing direct velocity data. Localization sensors like GNSS/INS, IMU, and wheel encoders tell the system where the vehicle is relative to detected objects.
99%
The system aims to handle 99% of scenarios encountered in deployment, with well-defined mitigation strategies for the remaining 1%.
Rather than stopping in uncertainty, Applied Intuition's stack uses learned models to produce probability distributions over actions. By fusing sensor outputs through ML inference, the planner evaluates thousands of candidate trajectories, planning around obstacles instead of halting in front of them.
Why this matters
This sensor fusion approach enables safer and more efficient autonomous mining operations, moving beyond simple GPS waypoint following to true perception-based navigation.
Terms in This Story
- Lidar
- A sensing method that uses laser pulses to measure distances and create detailed 3D maps of the environment.
- Radar
- A detection system that uses radio waves to determine the speed and distance of objects, especially moving ones.
- GNSS/INS
- Global Navigation Satellite System combined with Inertial Navigation System for precise positioning and orientation.
- ML inference
- The process of using a trained machine learning model to make predictions or decisions from new data.
Summarised from the linked release; details can be imperfect — always verify against the original source.